基于机器学习的油水层解释新方法--以新安边油田南部长7油层组为例  

A New Method for Oil/Water Layer Interpretation Based on Machine Learning:Taking Chang 7 Oil Layer Group in South of Xin’anbian Oilfield as an Example

作　　者：陶静 张宝辉 杨博[1] 周创飞[1] 胡晓雪 卜广平 路向伟[1] 屈乐[2] TAO Jing;ZHANG Baohui;YANG Bo;ZHOU Chuangfei;HU Xiaoxue;PU Guangping;LU Xiangwei;QU Le(No.6 Oil Production Plant,PetroChina Changqing Oilfield Company,Xi’an,Shaanxi 710200,China;Key Laboratory of Xi’an City for Tight Oil(Shale Oil)Development(Xi’an Shiyou University),Xi’an,Shaanxi 710065,China)

机构地区：[1]中国石油长庆油田公司第六采油厂,陕西西安710200 [2]西安市致密油(页岩油)开发重点实验室(西安石油大学),陕西西安710065

出　　处：《西安石油大学学报（自然科学版）》2023年第2期89-95,共7页Journal of Xi’an Shiyou University（Natural Science Edition）

基　　金：国家自然科学基金(52004223);陕西省创新能力支撑计划(2022PT-08);陕西省教育厅青年创新团队项目(22JP063)。

摘　　要：储层物性测井解释工作是石油勘探开发的一项重要任务。由于复杂的地质条件和沉积环境,储层非均质性与测井响应特征之间的非线性关系表明,线性测井响应方程和经验统计公式已不能有效表征储层特征,传统的储层特性解释方法与研究人员经验直接相关,存在一定程度的不确定性,因此本研究提出基于机器学习方法对新安边油田南部长7油层组进行油水层解释。首先,筛选测井序列齐全且有准确试油结果的数据作为样本数据,对其进行数据清洗;其次,基于一对一支持向量机(OVO SVM)与随机森林(RF)算法分别建立油水层解释模型进行对比分析,利用10折交叉验证进行模型参数优选;最后,利用参数优选后性能较高的OVO SVM模型对部分测井信息进行二次解释。研究结果表明,OVO SVM模型性能优于RF模型,解释准确率超过90%,对新安边油田新增数据采用该模型进行二次解释,变更解释结果层位54层。Logging interpretation of reservoir properties is an important task in petroleum exploration and development.Due to the complex geological conditions and depositional environment,the nonlinear relationship between reservoir heterogeneity and logging response characteristics shows that linear logging response equations and empirical statistical formulas can no longer effectively characterize reservoir characteristics.The traditional reservoir characteristic interpretation methods are directly related to the experience of researchers,and they have a certain degree of uncertainty.Therefore,this study proposes to interpret the oil and water layers of the Chang 7 oil layer group in the south of Xin’anbian Oilfield based on the machine learning method.Firstly,the data with complete logging sequence and accurate oil test results are selected as sample data,and the data are cleaned.Secondly,the oil/water layer interpretation models are established based on the one-versus-one support vector machine(OVO SVM)and random forest(RF)algorithms separately.The models are compared and analyzed,and the model parameters are optimized using 10-fold cross-validation.Finally,the OVO SVM model with higher performance after parameter optimization is used for secondary interpretation part of the logging information.The results show that the performance of OVO SVM model is better than that of RF model,and its interpretation accuracy is more than 90%.The model is used for secondary interpretation of new data in Xin’anbian Oilfield,and 54 layers of interpretation results are changed.

关 键 词：新安边油田 油水层解释 机器学习 模型优选 

分 类 号：TE34[石油与天然气工程—油气田开发工程]